BAD ENGINEERING IDEAS
Home • Created 2021/12/7 • Updated 2022/8/2 • Read time 17min • Discord
This page contains projects we have either already attempted or will try to attempt in
the future. I wanted to add content to this page like Bartosz Ciechanowski's but it's
clear no one will do what he does better than he does, so I have instead worked on
finding novel engineering solutions to problems or amusing experiments to run. Most of
this page is just for fun, so don't get too concerned about the space nukes section.
Computers for apocalypses. A fully mechanical, programmable, general computer that can
store, read, and write hundreds of lines of data, survive extreme radiation, a
thousand years in the dirt, and zero-G, all while fitting in the space of a bulky
laptop. This was partly inspired by projects like CollapseOS, Rosetta Code, looking
over the Antikytheramechanism, and a desire for hyper-efficiency from blogs like
What happens when solar flares wipe out all unshielded electronics? What happens if
the power grid is taken down for months at a time by China? What happens after a
successful EMP attack on your nuclear submarine or in your spaceship? How do you make
sure your computer will work after centuries of being placed on a shelf?
While I make no illusions that a fully mechanical general computer will outperform the
raw computational speed of a digital computer, we still need to significantly expand
on the analogue systems we completely abandoned fifty years ago for the sake of
surviving extinction-level events. A sentimental reason for wanting this could be
simply the ability to hold a physical instantiate of a program in your hands. It would
be the difference between holding a book and holding an e-reader, the e-reader will
always be dogshit compared to the book. The long-lasting nature of it interests me as
well. It would serve as an important artifact for future peoples to measure our
competence by, as you could dig it up from the ground and have it still be in working
order a thousand years later.
So how do we build this? Babbage's difference engine and analytical engine were
programmable, albeit in very limited fashion, but we want something more than a simple
MDAS system. It would likely have relatively limited memory, relatively limited bus
size, and so on, but enough to perform some otherwise advanced functions. We don't
want to give in to the advantages of a relay computer here since electro-mechanical
computers do not survive intense radiation bursts. We also don't want to make use of
what I would call "hyper-analogue" computation like the MoNIAC since those would fail
The good news is that almost every relevant mathematical operation you would care to
carry out is well understood and easily implimented as an analogue mechanism, as this
fire control computer shows. From single-mechanism MDAS'sAnalogue computation allows you to
have multiple operation types in the
same mechanism. E.g., addition and
subtraction, of integers or floating
points, can all be done with a single
rack and pinion mechanism. Similarly
with multiplication and division., mechanical integrators,
differential analyzers, or exact pinwheel calculators like the CURTA. We can even use
some neat tricks to make these functions more compact and accurate. Additionally,
almost all parts could be miniaturized using watch-making techniques, but to survive
being dropped or slammed, bulkier components in certain areas may be required. Either
way, we should be able to get all calculation outputs within two significant figures
without magnification and if near-microscopic laser'd markings are used in conjunction
with a magnifying glass over the output area, we could probably get calculation
outputs within five or more significant figures.
On top of math operations, there would also be a dedicated space for mechanical logic
operations. For inputs and programs, we could easily use bits/punches/sliders on solid
cards, which then code for which operations the mechanical computer would perform. If
card states code for choices of operations given a full instruction set, we now have a
Turing-complete mechanical computer. The cards would feed into the computer just like
paper ones would, but these cards are metal and the feed is manual. Let's say the
cards are not so clunky or large, we could probably fit fifty instructions on a 5x3
The interesting bit about a fully manual computer like this is that just as a program
can go forward in sequence, it can also go backwards in sequence by simply flipping
the card around. You could step back through a program and reverse all operations.
This means program cards double in function, as you can take known outputs and reverse
the calculation to derive the inputs those outputs resulted from.
Whatever the design we come up with, someone could use the same exact design but with
better materials, meaning as long as the design is good then the wear on the materials
is not a concern. Most of the gears would not have to come into contact with each
other, you'd only be engaging one mechanical operation at a time and moving the
mechanical output value of those operations between mechanisms, so the torque/binding
would be incredibly low.
This will probably never actually be done, but the fact that it's possible makes it a
juicy enough target that the members of Diogenesis might end up tackling it one day.
We made a cosmic ray catcher and used it to gamble. Or at least we tried to. High
energy particles from super novas sometimes slam into ram cards and flip a 1 to a 0.
Odilitime wrote a script for me that detects these kind of bitflips in allocated
memory space and we were betting on which half of the space a ray would pass through
first. The space was divided into two equally sized sections A and B, giving each a
50/50 chance of getting hit first. We even had a fancy image system that would swap
out the alien head on snerx.com's landing page with a new image indicating if A or B
IBM estimates that about one bitflip by cosmic ray happens per 256mb of data per
month, so we allocated 500mb of ram to the cosmic ray catcher process and waited for a
bitflip, expecting to see one in ~15 days. After three months with no observation we
realized that modern ECC fully accounts for cosmic rays and would never allow us to
observe the bitflip, which is really something we should have known beforehand but we
were so caught up in the joke of taking a lofty and beautiful idea like stars
exploding and using it for something base like gambling that we overlooked this
obvious problem. We had ten friends betting on this cosmic ray catcher and very
anti-climactically had to resolve the bet with a coin toss instead. It landed on
tails. We will probably revisit this project in the future so we can actually use
cosmic rays to gamble.
Using CIA research to generate visual binaurals. The famous-in-some-circles
documentation the CIA boldly hosts in their virtual reading room on their research
into a collection of projects they called the Gateway Process illuminates a great
number of things, primarily that the CIA fully believes, and has invested millions of
dollars into, parapsychological spirit magic, but in a very serious and scientifically
reproducible way that allows you, the layman, to astrally project too! This is only a
half-joke because clicking the link above will make it very clear the CIA takes this
very seriously, and further that very well known physicists and neuroscientists were
either contracted directly or had their research used to carry out the Gateway Process
This is interesting for many other reasons as well, but what we in the Snerx project
team have focused on is in trying to reproduce what the CIA calls a 'frequency
following response' from 'hemi-sync', now known more commonly as a binaural beat. The
CIA's documentation doesn't say anything controversial in this regard (even though it
may have been controversial at the time of their research many decades ago), so it
wouldn't be very worthy to try to reproduce this with audio - there are many videos on
YouTube where you can hear and experience a binaural beat. Instead, what we want to do
is see if we can generate a frequency following response that is trigged from visual
phenomena instead of auditory phenomena.
The way we plan on doing this is by taking two screens whose light is fully divided
between each eye with no overlap, like in a VR headset, and modulating the refresh
hertz and color projection to be within 30hz or lower of each other between the two
eyes. We will try refresh hertz first, with one eye's screen running at 30hz and the
other eye's screen at 60hz (or some similar combination), and if no response is
observed, then we will try the color frequency differentials to see if that has an
What we expect to observe is a visual binaural akin to the auditory binaural
experienced in a binaural beat, but instead of hearing the auditory illusion of a beat
produced as a response to the hemispheres of your brain synchronizing the auditory
stimuli, you should see the visual illusion of flashes produced as a response to the
hemispheres of your brain synchronizing the visual stimuli.
If we cannot find a way to produce a visual binaural in the same capacity that we can
easily produce an auditory binaural, this would still be a useful experiment since it
would tell us there is some special capacity that auditory sense has over the
hemispheric synchronization of your brain in which visual sense does not.
A reified halting problem. Turing's famous halting problem has been proven undecidable
but this was done a priori with logic rather than any kind of physical test or
empirical instantiate. A Turing machine running the program described in the halting
problem is something we could actually make, and further we could make two of them
with one seated inside the other just the same as Turing's description and then start
their feeds and have one feed the other as described in the original description.
What is interesting about this is that as soon as you start to describe this system
physically, it becomes apparent that it is likely to behave differently than the
purely abstract form of the system is concluded to behave as, since the physical
system will in fact either loop infinitely or immediately halt. It is important to
note that the physical looping would not be a recursive contradiction like in logic,
since logical contradictions do not physically exist.
I personally believe this will also help quasi-empirically (+) demonstrate a few other
tangentially related things in mathematics, namely that second-order-logic is invented
and not discovered and that Gödel's incompleteness is purely a symptom of math and not
Despite requiring very little money to do this physical test of the halting problem, I
have not found a single person who has attempted this on the internet. I am probably
just searching the wrong terms or something but really it should be totally trivial to
return dozens of pages of this exact thing and yet when searching for it not a single
result is immediately available. Even if other people have done this, it's worth doing
again for the reasons stated in the last paragraph.
Exotic solutions to warfare in space. Joe Halderman's The Forever War is the greatest
book on the difficulties of combat in space I've found and the main components of the
novel aren't even about that. Halderman's sci-fi novel allusion to the Vietnam war
presciently describes what the real enemy in space is - heat. The biggest concern with
space warfare is not an armed opponent, but heat dissipation.
Unintuitively, it's really difficult to rapidly cool things in space and any kind of
weapon system that generates heat (including your own body) needs to radiate the heat
away faster than you fire the weapon or it will die, melt, and explode, and not
necessarily in that order.
Heat transference happens three ways: convection, induction, and radiation. Convection
is when free atoms like gases or liquids bump into an object and transfer some of its
energy, which doesn't apply to space since space is a vacuum. Induction is when heat
energy transfers between atoms that are bonded together like heat moving down a metal
rod, and this doesn't work for most things in space either because you usually want
the heat to leave the system entirely instead of being trapped in it and having a
run-away heat problem. So all that's left for cooling things in space is radiation,
where EM forces allow atoms to radiate energy away without having to directly transfer
it to other atoms. The problem here is that this is generally much slower than the
other two options.
For any kind of space combat, with any kind of weapon system, you will be generating a
lot of heat. Radiating that heat away would require large radiation fins (like the
kinds you find on the sides of collocations) or something similar, and even then you
still can't radiate heat faster than you would be generating it by using the weapon
system, so what happens is you fire a few times and then have to wait for the weapon
to cool down before you can fire it again. Lots of video games have mechanics like
this for sci-fi railguns or similar weapons, and these games tend to accurately
capture how annoying it is to use the real-life counterpart since you get very few
uses of the weapon system before having to wait for cooldown in most cases.
The solution I came up with is to just use conductive/convective transference to pipe
the heat directly into the payload being delivered. So you just say fuck it to
mitigating the heat imbalance and use it to your advantage by throwing it into the
munitions you're going to send to the target anyways. This works in an obvious way with
missiles since the missile can get as hot as you want, it won't matter if they melt as
long as they are on the right path and far away from the firing system by the time
they explode. Conductive and convective heat capture are also useful for optimizing
sterling engines - something potentially useful for internal mechanical systems of the
vehicle carrying the payload.
Less obviously, this can also be used for temporary cloaking. True cloaking of a large
craft is broadly considered impossible by most astrophysicists and mil tech designers
because whatever wavelength of light you redirect typically amplifies the wavelengths
around it. E.g., blocking visible light generates more infrared light since retaining
visible light is adding energy as heat and heat is emitted as infrared. However, a way
around this is to use it to your advantage like the weapon system I limned in the last
paragraph. Yeah, you're going to be absorbing/generating a lot of heat, so just throw
it into a metal or dense-material cube until the material cannot load any more heat
without losing structural coherence and then eject the ultra-hot mass like a plasma
flare or like a squid inking to run away.
Doing it this way means you could technically absorb all light wavelengths across the
theoretically observable spectrum and otherwise be completely invisible for however
long the internal mass you're piping the heat to can maintain integrity (after which
it would begin radiating more heat back than it can take in). There is also a ratio
equation that could be derived where you can exactly calculate the max sustainable
time of your cloak by relating the overall surface of your craft to the heat
capacitance and mass of the object you're piping the craft's surface energy into.
The only way you'd be visually detectable is when you eject the super-heated mass, to
which you only need to clear a few seconds before being 'cloaked' again. You could
also be spotted if flying close to a craft and between any stars it's observing, as
you would be blotting out those stars. If you don't maintain radio silence you would
be detected, much like a submarine is only stealthy when running silent.
This is all to say that warp drives are real, aliens have already visited, it's absurd
to think you need to pay taxes, no resource is ultimately scarce, the universe has no
edge and neither does your mind, for to place a limit on thought is to think both
sides of the limit. This world is abundantly expansive and inexhaustible; you will
never run out of ways to amuse yourself or worthy ideas to explore so long as you task
yourself with doing so. Get excited, be energized, do more.
Three insights from an engineering dream. I've mentioned on /skell that I often use my
dreams to help me solve real mechanical problems. As described on that page, I do
roughly half of my work while fully asleep at night. I had a dream this morning that
gave me three insights into space engineering.
1. Scifi has lots of memes about zero-G vacuum welding, but for any substantial
infrastructure built in/for space, we need a complete zero-G vacuum manufacturing
line, from start to finish. This would also obviate problems like the ignition of
atmosphere since welding/manufacturing outside your pressurized hull means it is
2. Both fission and fusion reactors, but primarily fusion, are not just power sources,
but double in functionality as the precursors to atomic assembly - from dirt you can
create gold - which means they will be a critical component of atomically-precise
assembly and the crucial technology required for making all elements non-scarce.
3. There's an interesting distinction between mechanical artificial gravity and warped
artificial gravity. Mechanical AG is like the gravity wheel (or the 'cigar') portrayed
in most scifi conceptions, but this only gives you gravity in a very narrow area and
requires a very large moving hull. Lots of things can break there and it would be
relatively difficult to fix. Additionally, a mechanical AG would need to be so large
(at least 1km in diameter) to maintain a uniform sense of gravity at your feet as well
as your head, that the materials cost and assembly time becomes quickly infeasible if
you want to make more than one of them. Warped AG on the other hand only requires the
ground to be moving up at 1G of force inside a warp bubble with the warp field moving
at 1G downwards. This warped AG creates a uniform 1G of downward force across all the
space inside the warp bubble no matter how big or non-wheel-like the space is, and
there are no moving parts. Additionally, it would appear to outside observers as if
the space inside was staying still relative to them despite the fact that you are
technically flying upwards inside the bubble.
Social network engineering. The former CTO of Coinbase wrote an article on how Elon
Musk could effectively extort Twitter's staff by co-opting the userbase against them
as described in this article. The idea is neat but I wonder why this can't be done by
a third-party instead?
Someone independent of the Twitter deal could set up a secondary site and offer some
coin for every user that posts a 'final' proof-of-exit post exactly as described in
the linked article. Further, the more followers the account had, the more coin they
could get, incentivizing the most active users to switch platforms. Why shouldn't we
hold large social media platforms hostage like this? Honest question.
Any platform that put up bounties like this against Twitter, or Instagram, or whatever
other network would be able to hijack that network's userbase. Disallow advertising,
force any recommended content to link to educational content from museums and
universities, and you have effectively reset social media over night.